2022-RA-604-ESGO Urine and vaginal cytology detects endometrial cancer in women with postmenopausal bleeding

Introduction/Background

Postmenopausal bleeding (PMB) prompts urgent investigation with sequential invasive and costly tests that can be painful or distressing. A simple, non-invasive test to identify cancer and safely reassure women with benign causes of PMB would transform patient care. We previously showed proof-of-concept that malignant cells can be detected in urogenital samples of symptomatic endometrial cancer patients. Here, we aimed to prospectively validate the diagnostic test accuracy of urogenital cytology for endometrial cancer detection in women with PMB.

Methodology

In this prospective, multicentre diagnostic accuracy study, consecutive eligible women provided a self-collected voided urine sample and a Delphi screener clinician-collected vaginal sample before undergoing routine clinical investigations. Samples were assessed by two independent cytologists blinded to cancer outcomes. Discrepancies were settled by consensus review. Results were compared to standard clinical investigations and hysterectomy histopathology.

Results

Of 1864 participants, 115 (6.17%) had endometrial (n=99) or other pelvic malignancies (cervix-7, ovary-3, leiomyosarcoma-2, bladder-1, colorectal-2, metastatic pancreatic-1). The sensitivity and specificity of urogenital cytology for endometrial or any pelvic cancer detection were 80.8% (95% CI:71.7–88.0%) and 92.6% (95% CI:91.2–93.8%), and 80.0% (95% CI:71.5–86.9%) and 92.6% (95% CI:91.2–93.8%), respectively. The negative predictive value was 98.8% (95% CI:98.2–99.3%) for endometrial cancer detection and 98.6% (95% CI:97.9–99.1%) for any pelvic cancer detection. Of the 19 endometrial cancers missed by urogenital cytology, 2 (10.5%) had high-grade histology and 1 (5.3%) was ≥stage-II, meaning that cytology detected 95.8% of aggressive histology and 96.4% of locally advanced/metastatic cases. Seventeen of the missed cases (84.2%) were identified following unblinded cytology review of the sample, suggesting that natural tumour shed is ubiquitous in symptomatic endometrial cancer patients but current technology limits its clinical application.

Conclusion

This novel endometrial cancer detection tool holds great promise. Artificial intelligence solutions to screen cytology slides for rare malignant cells may improve its diagnostic accuracy.